mkn.gpu

CUDA/HIP C++20 convenience wrappers

======

Whether you are using CUDA or ROCM, we attempt to deduce from available headers.
If automatic detection fails, specify appropriate define like `-DMKN_GPU_CUDA=1`
See: inc/mkn/gpu/defines.hpp

Compile argument switches

Key             MKN_GPU_CUDA
Type            bool
Default         0
Description     activate CUDA as impl of mkn::gpu::*

Key             MKN_GPU_ROCM
Type            bool
Default         0
Description     activate ROCM as impl of mkn::gpu::*

Key             MKN_GPU_FN_PER_NS
Type            bool
Default         0
Description     expose functions explicitly via
                   mkn::gpu::hip::*
                   mkn::gpu::cuda::*

Key             _MKN_GPU_WARP_SIZE_
Type            uint
Default         use manufacturer provided (eg warpSize), usually 32
Description     override use if defined

Key             _MKN_GPU_THREADED_STREAM_LAUNCHER_WAIT_MS_
Type            uint
Default         1
Description     Initial wait time in milliseconds for polling active jobs for completion

Key             _MKN_GPU_THREADED_STREAM_LAUNCHER_WAIT_MS_ADD_
Type            uint
Default         10
Description     Additional wait time in milliseconds for polling active jobs for completion when no job is finished.

Key             _MKN_GPU_THREADED_STREAM_LAUNCHER_WAIT_MS_MAX_
Type            uint
Default         100
Description     Max wait time in milliseconds for polling active jobs for completion when no job is finished.

Key             MKN_CPU_DO_NOT_DEFINE_DIM3
Type            bool
Default         false
Description     if true, skips defining dim3 struct which is usually provided by gpu headers